Last Tuesday Morning before the HIMSS conference got going in New Orleans, Healthcare Informatics' editors had a chance to sit down over breakfast with our editorial advisory board to get some feedback on the types of stories they would like to see more of in our pages and online. One topic several of them brought up was the importance of data governance to strategic initiatives such as meaningful use, ICD-10 compliance, and data warehouses. We agreed to write some in-depth case studies on data governance in the near future.
Not two hours later I found myself sitting in several educational sessions in which data governance played a key role. First, Gregory Veltri, CIO for the Denver Health and Hospital Authority, Colorado's largest safety-net healthcare system, talked about his experience helping the organization build a data warehouse. He said that although meaningful use may feel at times like a reporting burden, it is providing a framework to do much more. “It allows me to ask, how can I extend our data model to give physicians best info about patients, including genomic data for predictive analytics.”
Over several years, Denver Health’s data warehouse has built up 27 interfaces from every clinical application. Veltri talked about the several steps involved in data management:
1. Create a common data dictionary of defined measures. (You have to have one definition for every data element you have. If not, you get hopelessly lost, Veltri said.)
2. Collect data (Unstructured data is the toughest).
3. Measure and analyze.
4. Make the data actionable.
5. Build a predictive analytics model.
Veltri said one challenge is that the process takes time and sponsors may get antsy because they aren’t seeing results right away. “They may say, ‘we are 18 months into this $4 million project and we’ve got nothing to show for it. Should we stop?’”
The goal of the effort must be to ultimately predict the behavior of the system and its consumers, and to drive down costs while improving outcomes. Users can come up with meaningful insights, and think about their work in a new way. This can drive innovation in your organization. For instance, you can show a breakdown of the cost of diabetic care by clinician, by clinic, and by practice and allow providers to drill down and look for reasons behind the differences.
Veltri says the most important thing to remember is that technology alone will not resolve data governance and quality issues. “This requires a cultural shift in how data is perceived and managed,” he said.
In a separate session, Paula Edwards, a partner with consulting firm HIMformatics, agreed with Veltri that if these efforts are led and solely pushed by IT, they fail. “If data governance is viewed as an IT project, it is not going to be a long-term program. It has got to be driven by the business and clinical side.”
Once you clarify who is responsible for which parts of the project and provide transparency into the decision-making process, Edwards recommends focusing on data quality, integration, master data management, standards and metadata.
It is important to automate and streamline business rules for integrating and cleaning data from different sources. Standards should be set for tool sets and policies for data management. “Master data management” refers to defining a single source of truth, including standard codes and descriptions. Data dictionaries need to be created encompassing both business and technical definitions and data lineage. Data stewards should be assigned, she said. They are business owners who provide expertise on what data mean involving a particular subject.
Edwards said it is important to do outreach and education on these principles. “Self-service business intelligence tools don’t help very much if users don’t understand what the data means,” she said. To keep the effort visible, she suggests leveraging existing governance meetings. “Add data governance as an item on their agenda,” she said. Some new workgroups may have to be generated at the tactical level. But data governance should start to be engrained as part of the organizational culture.”
“When I go in to work with organizations, it is shocking to me that no one realizes how much time analysts spend manually pulling data together,” Edwards said.It is important to reduce costs and complexity of data-related initiatives, she added. Data governance can allow you to avoid bad decisions based on incomplete or inaccurate data.